Lavender and Regie.ai: Two Sales-Coaching Agents, Two Opposite Unlock Thresholds

Two agents in the same procurement bucket — "AI sales coaching" — bill on opposite curves. Lavender charges $49 per user per month for the first 25 seats, then steps up to $89 per user above 25 (yes, the per-seat price gets more expensive as you scale; we will get to that). Regie.ai charges $5,000 per month flat. At a 5,000-employee SaaS company with 300 active SDRs and AEs on Outreach and Salesloft, one of these never needs an unlock calculation. The other lives or dies by one.

This post is about which is which, and why the answer is not the one a sales rep would tell you.

The two pricing shapes, with the math written out

Lavender is a per-user agent. The catalog entry shows costPerUser = $49, enterpriseCostPerUser = $89, enterpriseThreshold = 25. Read that carefully: the per-seat price goes up once you cross 25 users. This is unusual — most enterprise pricing ladders tier down with volume, not up — and it is deliberate. Lavender's product surface (real-time email coaching inside Gmail and Outlook) is priced to make small-team adoption cheap and to capture enterprise budget for genuine deployments. There is no flat-fee equivalent. You cannot "unlock" Lavender; the agent costs $89 × seat_count per month above 25 reps, every month, forever.

For a 300-rep deployment that is 300 × $89 × 12 = $320,400/yr in agent cost. There is no setup floor — per-user agents return $0 from effectiveSetupCostUsd by design.

Regie.ai is a flat-fee agent. The catalog row shows monthlyCost = $5,000, no costPerUser, no enterpriseThreshold. That means $60,000/yr for the agent itself plus a setup cost computed by the engine as max($15,000, monthlyCost) = $15,000. Year-1 all-in: $75,000. Year-2 onward: $60,000.

These are not comparable price tags. They are two different shapes of cost curve, and an enterprise CFO needs to recognize which shape they are buying before scoring either one.

What each agent actually compresses

The catalog assigns compression percentages per (agent, target tool) pair. Lavender has none. Zero AgentToolImpact rows. It sits in the agent table as a tool you can buy, but the engine does not credit it with reducing seats on Outreach, Salesloft, Salesforce, or Apollo.io. That is not an oversight — it is a category statement. Lavender coaches the rep writing the email; it does not deflect the seat the rep occupies in the sales engagement platform.

Regie.ai is the opposite. Its impact rows: Salesloft 45%, Outreach 45%, Apollo.io 43%, Salesforce 30%. Those two sales-engagement numbers are the load-bearing ones. They say the agent — which generates sequences, personalizes touches, and books meetings autonomously — is credited with displacing close to half of the per-seat headcount on both engagement platforms for the population it covers. Salesforce gets less (30%) because the CRM seat carries account-ownership and reporting work the agent doesn't absorb; the catalog treats the cadence layer as the displaceable surface.

This is the asymmetry the procurement team needs to see on one slide. Lavender's value lives entirely in productivity-per-rep — faster reply rates, better book-to-meeting ratios, less time spent staring at a blank message field. None of that shows up in a seat compression model because none of it removes a Salesloft license. Regie.ai's value is, by the catalog's accounting, mostly seat displacement: it lets a smaller SDR team cover the same outbound volume.

If you are buying for productivity, the two questions are different. If you are buying to retire SaaS spend, only one of them produces a number you can put in a renewal email.

Why Regie.ai needs an unlock threshold and Lavender does not

This is where the flat-fee shape bites. A flat-fee agent has to clear its own annual cost before the engine marks it profitable. The viability rule in the analysis engine: netAnnualUsd = annualGrossSavings − annualAgentCost, where annual cost is monthly × 12 + setupCostUsd. For Regie.ai that floor is $75,000 in year one, $60,000 thereafter.

The gross savings come from compressionPct × active_seats × (monthly_cost / total_seats) × 12 on every targeted tool, with the MAX-overlap rule resolving conflicts when other agents (AiSDR, Artisan Ava, Avoma) also target the same SaaS row. Among Regie's own four tools, Salesloft is the biggest single contributor — Salesloft list is $125/seat/mo, the highest per-seat cost, so its 45% rate moves the most dollars even though Outreach carries the same 45%. (One caveat the MAX rule forces: Artisan Ava sits at 63% on Salesloft and Outreach in the catalog, so if Ava is also in scope the engine credits Ava's higher rate on those two tools and Regie's contribution collapses to Apollo and Salesforce. Regie wins the Salesloft lever only when it's the agent you actually deploy there.)

How many active Salesloft seats does it take to unlock Regie.ai standalone? Treat Salesloft as the only contributor and solve:

annual_gross_savings ≥ $60,000  (Year 2)
seats × 0.45 × $125 × 12 ≥ $60,000
seats ≥ 89

That is the floor if you already own enough Salesloft seats. In year one with the $15K setup amortized, the threshold pushes to ~111 Salesloft seats. Layer in Outreach at 45% × $100/seat/mo and the combined sales-engagement footprint Regie needs drops to roughly 90-130 active reps across the two tools, depending on the split — the higher Outreach rate pulls the combined unlock down meaningfully from the Salesloft-only floor. Below that, the engine returns unprofitable and the action plan slot stays empty.

Lavender has no such threshold because there is nothing to amortize. A 30-rep team paying 30 × $89 × 12 = $32,040 either values the coaching enough to justify the line item or does not. The decision is qualitative, not viability-gated. This is the structural advantage of per-user pricing for the vendor: every rep you add is incremental revenue. It is also the structural ceiling for the buyer: there is no economy of scale baked in. A 300-rep deployment pays linearly. A 3,000-rep deployment pays linearly. Lavender will never appear in a CFO's "top compression wins" list because compression is not what it does.

The mechanics behind the unlock math are the same ones we walked through in the AiSDR vs Salesforce post — flat fee divided by per-rep gross savings yields a hard floor in headcount. Same shape; different agent, different target stack.

Where the two agents collide: the same SDR's Outreach license

This is the interesting question for an enterprise running both. Suppose the rev-ops team pilots Lavender on the AE team and Regie.ai on the SDR team. They share Outreach. The MAX-overlap rule kicks in: whichever agent has the higher compressionPct on a given (rep, Outreach) intersection wins.

Lavender's contribution to Outreach compression is 0 (no impact row). Regie.ai's is 45%. So on every SDR seat that Regie covers, the engine credits 45% Outreach displacement. On every AE seat that only Lavender covers, the engine credits 0% — because Lavender does not compress Outreach by the catalog's accounting.

This produces a counterintuitive procurement conclusion: deploying Lavender to AEs does not shrink your Outreach renewal anchor. Deploying Regie.ai to SDRs does. If the CFO's question is "what gets us a smaller Outreach contract next year," the answer is Regie, not Lavender, even though Lavender's per-rep price is half. The savings math runs through compression, not through agent cost.

For Salesloft the gap is just as wide. Regie.ai compresses Salesloft 45%; Lavender 0%. On a stack where Salesloft is the higher per-seat anchor ($125/mo vs Outreach $100/mo), Regie.ai is the lever that moves the renewal even before you add the Outreach 45% on top. Buying both is fine. Believing both reduce your SaaS spend is not.

The four-agent crowding problem

The sales-coaching category is one of the most saturated AgentToolImpact intersections in the catalog — and the rates are no longer clustered, which makes the MAX rule decisive. Four agents target Outreach, ranging from Avoma at 20% and AiSDR at 30% up to Regie.ai at 45% and Artisan Ava at 63%. A fifth — 11x Alice — sits adjacent in the same sales-engagement bucket via Salesforce 30% and Salesloft 45%, but does not credit Outreach compression in the catalog. On Apollo.io the spread runs AiSDR 20%, Regie 43%, Artisan 60%. Six agents target Salesloft, from Instantly at 18% and Bland AI at 20% through AiSDR 30%, 11x Alice 45%, and Regie.ai 45%, topping out at Artisan Ava at 63%.

The MAX-overlap rule means buying three of these does not stack into a bigger number — the engine takes the single highest rate and discards the rest. On Outreach that's Artisan Ava's 63%; on Salesloft it's also Ava's 63%; adding Regie or AiSDR on top of Ava produces zero additional compression credit on those tools. The procurement implication: each marginal agent in this category is a redundancy purchase, not an additive one. You are paying for product surface area (UI, integrations, voice tone) — not for incremental seat displacement.

The cleanest way to play this category at enterprise scale: pick one flat-fee agent for the seat compression thesis and layer Lavender or Sybill on top for the productivity surface area that does not show up in compression math. On raw catalog rates Artisan Ava now tops the engagement-platform leaders (63% on both Salesloft and Outreach), so if maximum Salesloft/Outreach compression is the only goal, Ava is the MAX-rule winner; Regie.ai (45% on both, plus Apollo 43% and Salesforce 30%) is the broader-stack play when you want one agent spanning cadence, data, and CRM, and 11x Alice is the pick when the anchor is Salesforce-and-ZoomInfo heavy. Whichever single flat-fee agent you choose, the second flat-fee agent in the bucket is redundancy. Two distinct line items in the budget — the compression agent and the productivity agent — two distinct justifications, no double-counting in the savings model.

This same crowding pattern is why we wrote the case for sourcing every compression percentage — when four vendors all claim 30% on the same SaaS row, the catalog's job is to pick the highest defensible number, not sum them.

Worked example: 12,000-employee SaaS company, 240 active reps

Take a 12,000-employee SaaS company with the following sales stack:

Sales-stack ACV: ~$1.04M.

Scenario A: Lavender across all 240 active reps. Cost: 240 × $89 × 12 = $256,320/yr. Compression credited by the engine: $0 (no impact rows). Year-1 net seat savings: $0. The productivity case has to stand on its own — closed-won lift, reduced ramp time, reduced manual sequence editing. Real, but not in this ledger.

Scenario B: Regie.ai standalone, same stack. Annual cost: $5,000 × 12 + $15,000 = $75,000 year 1; $60,000 year 2.

Gross compression savings (engine math, MAX-overlap, pricing-model gate, all per_seat so no zeroing):

• Salesloft: 165 × 0.45 × $125 × 12 = $111,375/yr
• Outreach: 235 × 0.45 × $100 × 12 = $126,900/yr
• Apollo.io: 175 × 0.43 × $49 × 12 = $44,247/yr
• Salesforce: 240 × 0.30 × $100 × 12 = $86,400/yr
• Gross annual: ~$368,922

Year-1 realistic (deploy_agent realization factor 0.4): $368,922 × 0.4 = $147,569 realized minus $75,000 agent cost = +$72,569 net year 1. Year 2 net (no setup, same realization): +$87,569.

(This assumes Regie is the agent deployed on Salesloft and Outreach. If Artisan Ava is also in scope, the MAX rule reassigns those two tools to Ava's 63% and Regie's credited contribution drops to Apollo + Salesforce — the engine never double-counts the overlap.)

By year three, with realization closer to steady-state and renegotiation actually closing on the Salesloft/Outreach renewals, the math compounds further — but the year-1 net is the number the procurement team should anchor on for the initial business case.

Scenario C: both, on different rep populations. Lavender on 60 AEs: $64,080/yr. Regie.ai on 180 SDRs targeted at Salesloft + Outreach: $75,000/yr. Engine treats them as non-overlapping in compression credit (Lavender's 0% does not contend with Regie's 30-45%). Combined year-1 agent spend: $139,080. Combined gross compression savings: ~$368,922 (Regie's contribution; Lavender adds zero). Year-1 net after realization discount: ($368,922 × 0.4) − $139,080 = +$8,489.

That barely-positive number is the whole point: the bundle now clears year one on Regie's compression alone, with Lavender contributing zero to that line. The discipline still holds — put Lavender's productivity case (revenue lift, ramp time, win rate) into a separate business case rather than crediting it inside the seat compression model. The difference from the old catalog is that Lavender is now pure upside layered on an already-positive compression bundle, not the thing rescuing a negative. Mixing the two ledgers is still the most common error we see in flat-fee-plus-per-user sales stacks. We covered the same dynamic on the renegotiation side — the seat math and the productivity narrative belong on different pages of the deck.

The bottom line

Two agents, same category, opposite economic shapes, opposite procurement defenses.

Lavender is a productivity tool priced per user, with an inverted threshold that gets more expensive at scale. It will never appear in a seat compression analysis because it does not compress any seats by the catalog's accounting. Buy it on closed-won lift, not on SaaS renewal pressure.

Regie.ai is a seat displacement tool priced flat, with a hard ~89-111 Salesloft seat unlock floor and material credit on Outreach (equal 45%), plus Apollo.io and Salesforce. Buy it on the renegotiation anchor it creates at the sales-engagement-platform renewal.

If a procurement team is evaluating both side by side on a single spreadsheet column called "annual cost," the spreadsheet is wrong. They belong in different columns answering different questions. If both questions are real, buy both. If only the SaaS spend question is real, buy Regie.ai and skip Lavender. If only the productivity question is real, buy Lavender and skip Regie.ai. Do not buy either because the rep on the other end of the Zoom call mentioned the other one as competition — they are not.

Run your own sales stack through the SeatCompress calculator to see which agent in this category clears the viability threshold for your active rep count. The unlock floors are deterministic; the per-rep price tags are not the input you think they are.